A second order training algorithm for multilayer feedformward neural networks

摘要

According to the adjusting principle, we propose a second order training algo- rithm for multilayer feedforward neural network after analyzing the slow convergence rea- son of traditional BP algorithm. The better region of the learning factors for this proposed algorithm to achieve good learning performance is also given. A computer simulation ex- periment on XOR problem demonstrates the correctness of our theoretical analysis and the effectiveness of the proposed algorithm.